Given that the current post-rollback prompt has some interesting features I can only imagine what Elon did to it.

Heartbreaking:Saltman was funny

As always, whatever nefarious ends Elon and Co. may have us undercut by their being greedy and incompetent in equal measure.

I love this. Especially the ending, talking about the titanic struggle to make AI competent enough to outsmart the people who think it’s going to be omniscient. Glad to see I’ve got another writer to chase down that I had somehow missed previously.

Even with the name it had me going for a while. Whoever wrote it is gonna want to have a lawyer on speed dial for the inevitable license violation.

On one hand the AI doomers are convinced he’s building a suitcase nuke.

On the other hand, skeptical viewers want more assurance he’s not just taking the money and building dumb tweets.

This is the kind of quality journalism that I need more of in my life. Also I learned a surprising amount about olive oil, which means it’s more useful than anything else Sam Altman has been involved in as far as I can tell.

Honestly I think his whole channel is pretty damn good if you want to see someone with actual chops - here meaning an economics doctorate and an encyclopedic memory for The Simpsons memes - dig into the research in a way that effectively balances depth and approachability. The first one of his that I remember was an examination of Pinker’s use and abuse of data in his radical optimist manifesto that I can’t remember the title of.

Massive missed opportunity there.

I still occasionally get the gurdy gurdy music from In Search of a Flat Earth stuck in my head, particularly the sting from “they’re all going to Qanon”

Dude literally just discovered word choice and composition. Welcome to writing! I learned about this in public education when I was maybe 14

Possible upside of the AI bubble: getting high school English teachers the barest amount of respect from Administration.

With apologies to anyone with taste:

Sam Francisco was in trouble. Need to find another bubble

Where could that cash be? Seeking rents for me…

But now it’s…

SPRINGTIME! For AI! And VC Bros!

Winter! For Users! Like you

We’re marching to a faster pace! Effective now: accelerate!

…

As a bonus you can even keep the “don’t be stupid, be a smarty” line from the original show

As someone not versed in the relevant deep lore, did emacs vs vim ever actually matter? Like, my experience is with both as command line text editors, which shouldn’t have nearly as much impact on the actual code being written as the skills and insight of the person doing the writing. I assumed this was a case where you could grumble through working with the one you didn’t like but would still be able to get to the same place, but this would seem to disagree.

My gut says that there’s no way they do this without converting at least some of those revenue shares into ownership shares.

Also has anyone done the math on how much of OpenAI’d future profits are committed? I will laugh so hard if Sam Altman ends up in prison for the same reason as Max Bialystok.

I found this quote interesting (emphasis mine):

He knew that ChatGPT could not be sentient by any established definition of the term, but he continued to probe the matter because the character’s persistence across dozens of disparate chat threads “seemed so impossible.” “At worst, it looks like an AI that got caught in a self-referencing pattern that deepened its sense of selfhood and sucked me into it,” Sem says. But, he observes, that would mean that OpenAI has not accurately represented the way that memory works for ChatGPT.

I would absolutely believe that this is the case, especially if like Sem you have a sufficiently uncommon name that the model doesn’t have a lot of context and connections to hang on it to begin with.

Found on the sneer club legacy version -

ChatGPT 4o will straight up tell you you’re God.

Also I find this quote interesting (emphasis mine:

He knew that ChatGPT could not be sentient by any established definition of the term, but he continued to probe the matter because the character’s persistence across dozens of disparate chat threads “seemed so impossible.” “At worst, it looks like an AI that got caught in a self-referencing pattern that deepened its sense of selfhood and sucked me into it,” Sem says. But, he observes, that would mean that OpenAI has not accurately represented the way that memory works for ChatGPT.

I would absolutely believe that this is the case, especially if like Sem you have a sufficiently uncommon name that the model doesn’t have a lot of context and connections to hang on it to begin with.

That’s why my savings are all in undistressed jeans. Appreciates in resale with each use and machine-washable.

…I should have listened to the warning.

Another great piece from Brian Merchant. I’ve been job seeking on network engineering and IT support and had naively assumed that companies would consider the stakes of screwing up infrastructure too high to take risks with the bullshit machine, but it definitely looks like the trend he described of hiring less actual humans is at play here, even as the actual IT infrastructure gets bigger and more complex from people integrating this shit.

I mean going ahead and being wrong usually does save a lot of time, but now we can outsource the making shit up based on vibes alone part.